Learning internal representations by error propagation
Parallel distributed processing: explorations in the microstructure of cognition, vol. 1
Computer as Thinker/Doer: Problem-Solving Environments for Computational Science
IEEE Computational Science & Engineering
//ELLPACK: a numerical simulation programming environment for parallel MIMD machines
ICS '90 Proceedings of the 4th international conference on Supercomputing
Parallel Computing: Emerging from a Time Warp
IEEE Computational Science & Engineering
PYTHIA: a knowledge-based system to select scientific algorithms
ACM Transactions on Mathematical Software (TOMS)
Software architecture of ubiquitous scientific computing environments for mobile platforms
Mobile Networks and Applications - Special issue on mobile computing and system services
ACM Transactions on Mathematical Software (TOMS) - Special issue in honor of John Rice's 65th birthday
MultiAgent System Support of Networked Scientific Computing
IEEE Internet Computing
From Scientific Software Libraries to Problem-Solving Environments
IEEE Computational Science & Engineering
Computational science, mathematics and software
Hi-index | 0.00 |
Many scientists and engineers could be using computation in their work to great advantage but aren't. Why is this? Well, largely for the same reasons that even some experienced computational scientists aren't yet using wavelets, or neural networks, or perhaps parallel programming, even if these methods could help them. They simply might not know very much about the techniques, and they're often too busy to learn. This points up the need for problem-solving environments, or PSEs--high-level systems that give a worker in some application area comprehensive and powerful computational assistance while hiding the low-level detail as much as possible. Purdue is a center of research on how to build these tools of the future, some of which are already in use. The authors describe Pythia, an "intelligent assistant" they are incorporating into several PSEs that are designed for solving partial differential equations. Pythia uses a knowledge base to help choose the methods, grid and other parameters, and machines best suited to solving a particular PDE problem. To do this, problems must first be grouped into categories with similar characteristics, and matched against similar problems in the database for which good solution methods are known. The abilities of various neural-network methods to do this classification are compared here. Feedforward multilayer perceptrons trained with enhanced variations of backpropagation, and a neuro-fuzzy method as improved by this team, did especially well and outperformed a traditional non-neural technique.